MinMo: A Multimodal Large Language Model for Seamless Voice Interaction

Recent advancements in large language models (LLMs) and multimodal speech-text models have laid the groundwork for seamless voice interactions, enabling real-time, natural, and human-like conversations. Previous models for voice interactions are categorized as native and aligned. Native models integrate speech and text processing in one framework but struggle with issues like differing sequence lengths and insufficient pre-training. Aligned models maintain text LLM capabilities but are often limited by small datasets and a narrow focus on speech tasks. In this work, we introduce MinMo, a Multimodal Large Language Model with approximately 8B parameters for seamless voice interaction. We address the main limitations of prior aligned multimodal models. We train MinMo through multiple stages of speech-to-text alignment, text-to-speech alignment, speech-to-speech alignment, and duplex interaction alignment, on 1.4 million hours of diverse speech data and a broad range of speech tasks. After the multi-stage training, MinMo achieves state-of-the-art performance across various benchmarks for voice comprehension and generation while maintaining the capabilities of text LLMs, and also facilitates full-duplex conversation, that is, simultaneous two-way communication between the user and the system. Moreover, we propose a novel and simple voice decoder that outperforms prior models in voice generation. The enhanced instruction-following capabilities of MinMo supports controlling speech generation based on user instructions, with various nuances including emotions, dialects, and speaking rates, and mimicking specific voices. For MinMo, the speech-to-text latency is approximately 100ms, full-duplex latency is approximately 600ms in theory and 800ms in practice. The MinMo project web page is https://funaudiollm.github.io/minmo, and the code and models will be released soon.

CosyVoice 2: Scalable Streaming Speech Synthesis with Large Language Models

In our previous work, we introduced CosyVoice, a multilingual speech synthesis model based on supervised discrete speech tokens. By employing progressive semantic decoding with two popular generative models, language models (LMs) and Flow Matching, CosyVoice demonstrated high prosody naturalness, content consistency, and speaker similarity in speech in-context learning. Recently, significant progress has been made in multi-modal large language models (LLMs), where the response latency and real-time factor of speech synthesis play a crucial role in the interactive experience. Therefore, in this report, we present an improved streaming speech synthesis model, CosyVoice 2, which incorporates comprehensive and systematic optimizations. Specifically, we introduce finite-scalar quantization to improve the codebook utilization of speech tokens. For the text-speech LM, we streamline the model architecture to allow direct use of a pre-trained LLM as the backbone. In addition, we develop a chunk-aware causal flow matching model to support various synthesis scenarios, enabling both streaming and non-streaming synthesis within a single model. By training on a large-scale multilingual dataset, CosyVoice 2 achieves human-parity naturalness, minimal response latency, and virtually lossless synthesis quality in the streaming mode. We invite readers to listen to the demos at https://funaudiollm.github.io/cosyvoice2.

Speech Corpora Divergence Based Unsupervised Data Selection for ASR

Selecting application scenarios matching data is important for the automatic speech recognition (ASR) training, but it is difficult to measure the matching degree of the training corpus. This study proposes a unsupervised target-aware data selection method based on speech corpora divergence (SCD), which can measure the similarity between two speech corpora. We first use the self-supervised Hubert model to discretize the speech corpora into label sequence and calculate the N-gram probability distribution. Then we calculate the Kullback-Leibler divergence between the N-grams as the SCD. Finally, we can choose the subset which has minimum SCD to the target corpus for annotation and training. Compared to previous data selection method, the SCD data selection method can focus on more acoustic details and guarantee the diversity of the selected set. We evaluate our method on different accents from Common Voice. Experiments show that the proposed SCD data selection can realize 14.8% relative improvements to the random selection, comparable or even superior to the result of supervised selection.

Data Augmentation based Consistency Contrastive Pre-training for Automatic Speech Recognition

Self-supervised acoustic pre-training has achieved amazing results on the automatic speech recognition (ASR) task. Most of the successful acoustic pre-training methods use contrastive learning to learn the acoustic representations by distinguish the representations from different time steps, ignoring the speaker and environment robustness. As a result, the pre-trained model could show poor performance when meeting out-of-domain data during fine-tuning. In this letter, we design a novel consistency contrastive learning (CCL) method by utilizing data augmentation for acoustic pre-training. Different kinds of augmentation are applied on the original audios and then the augmented audios are fed into an encoder. The encoder should not only contrast the representations within one audio but also maximize the measurement of the representations across different augmented audios. By this way, the pre-trained model can learn a text-related representation method which is more robust with the change of the speaker or the environment.Experiments show that by applying the CCL method on the Wav2Vec2.0, better results can be realized both on the in-domain data and the out-of-domain data. Especially for noisy out-of-domain data, more than 15% relative improvement can be obtained.

Transformer-based Online CTC/attention End-to-End Speech Recognition Architecture

Recently, Transformer has gained success in automatic speech recognition (ASR) field. However, it is challenging to deploy a Transformer-based end-to-end (E2E) model for online speech recognition. In this paper, we propose the Transformer-based online CTC/attention E2E ASR architecture, which contains the chunk self-attention encoder (chunk-SAE) and the monotonic truncated attention (MTA) based self-attention decoder (SAD). Firstly, the chunk-SAE splits the speech into isolated chunks. To reduce the computational cost and improve the performance, we propose the state reuse chunk-SAE. Sencondly, the MTA based SAD truncates the speech features monotonically and performs attention on the truncated features. To support the online recognition, we integrate the state reuse chunk-SAE and the MTA based SAD into online CTC/attention architecture. We evaluate the proposed online models on the HKUST Mandarin ASR benchmark and achieve a 23.66% character error rate (CER) with a 320 ms latency. Our online model yields as little as 0.19% absolute CER degradation compared with the offline baseline, and achieves significant improvement over our prior work on Long Short-Term Memory (LSTM) based online E2E models.